(and Nick)
Agreed. I wouldn't put much stock in Adobe's forums -- while there are a few people who are real gems / know a ton, most there know very little on a technical level. Unless someone from Adobe said "we're not supporting it", it's pure conjecture.
What's overlooked here is Adobe and Apple still have a contentious, love-hate relationship. Both need each other but both find ways to butt heads often, especially when Jobs was still with us. Over Flash, over OpenCL. Apple basically ****** on Flash in a very public way some time ago, instead making a vocal push for HTML 5 (which frankly was the RIGHT thing to do -- Flash is a lousy technology some levels and as everyone knows has to be constantly updated for security reasons). So what goes around comes around -- this lackluster support might be Adobe's "return slap" on some level (although both companies have new leadership since the Flash thing). I would expect Adobe to fully support OpenCL workflows only at such time as it improves the money in their pockets. Might be a while.
The other thing people need to understand about GPGPU (or GeneralPurpose (computingon) GPU) is that whether you use CUDA or OpenCL, it's a tricky thing to implement. It's not like they can flip a few switches or change a few lines of code around and everything works 10X faster. Rewriting even one part of an app to do this well, can absorb many developer hours. The problem is compounded by: a) substandard documentation of the current OpenCL API -- something that has visibly turned some developers off and is wholly due to Apple's inaction; b) drivers that are not yet well optimized even on the level of single GPU proceses (both AMD and Apple's fault); and c) you have to find the right places to do OpenCL in your app (no one's fault -- it's inherent to the technology).
It's not at all analogous to the old days when apps were switching over from PPC to Intel, where the whole app needed to be tweaked and recompiled, etc. This is more like some of the functions in your app will scream if you port them to OpenCL, and some will not improve at all and you've wasted your developer time. This is part of the reason why Adobe makes some functions and filters OpenCL/CUDA friendly, some CUDA only, and some work with only the CPU. The other part is I think they just don't give a **** because it's Apple's agenda and not theirs, and it's one more thing they can use as leverage to get whatever it is they want from Apple.
